Hi everyone,
I'm currently working on the splicing variants of a protein. For the analysis, I extracted variants from two databases in Fasta format. Before downstream processing, I want to make sure that no variant is repeated. So, how can I check this? I m using standalone BLAST on Linux. I tried using MegaBlast, but I'm not very confident about the results... in results e value is mostly zero and show 100 percent identity among sequences, bit score goes very high sometimes... the only thing which differ is query coverage... the start as well as end sites position vary in query and database sequence... I had to ask u... in the splice variant case can I make the query start and end difference as the basis of the difference of splice variants. I shall be waiting for ur response.
Mazhar Hussain
BLAST will not be a good tool for this. BLAST scores are based on both percentage identity and length of the matched sequence. You don't specify what type of database you are searching (your own built from cDNAs of the alternatively spliced mRNAs from one species; or GenBank, EMBL or other public databases of many species) or what your search query sequence is (the gene with introns; the longest of all the alternatively spliced mRNAs, etc).
Well these are cDNA sequences (spliced) in FASTA format from NCBI and ENSAMBL.
I am guessing that maybe you are doing dozens of BLAST queries, each with a differently spliced form as the search query, and then form the "hit" results you want a set of unique "hits" removing the duplicates. Are they all from the same species? Or are you interested for example in finding out that humans have spliced forms A, B and C while Gorillas have forms A and C but not B?
So you have FASTA files (the extracted splice variants) and want to rule out duplicates?
This is easily be done using either various tool-kits or clustering methods. As far as i know the EMBOSS package contains such a tool (skipredundant ), as also the FASTX-toolkit, PRINSEQ, and many many more. You can also consider to use clustering approaches like UCLUST or CD-Hit, which can be configured in such a way that they will cluster only sequences sharing 100% identity and also are similar in length (they will output also one representative for each such cluster and also provide information about the level of redundancy, i.e. a mapping file).
This task is very very common in sequence analysis. So searching for terms like 'deduplication', 'dereplication', or 'collapse' in combination with FASTA will give you hundreds of hits either to ready to use tool-kits or to customized solutions (based on various scripts languages like bio-python, bio-perl, ...).
you can also run unique.seqs() in MOTHUR. you will end up with a fasta file with only unique sequences and a names file telling you how often which sequence occurs.
command is: unique.seqs(fasta=FILE.fasta)
A problem with something like what MIna Bashir is suggesting, is that when you BLAST with a human gene, (say 50 kb with 17 exons interspresed along with introns that may or may not all show up n each message) the BLAST result will contain several "hits" for each mRNA/cDNA sequence in the database. The match to each exon will download as a separate sequence in the output. So assuming there are not SNPs or different alleles and all the different messages are done by alternative splicing a list of 70 different alternatively spliced variants should come out as just the 17 exons using a tool like UNIQUE.SEQS
Also, a tool like UNIQUE.SEQS will count sequences as different if they have a change in the length of poly-A tail or other differences that are not due to alternative spicing.
It sounds like what you really want is to align each mRNA to the complete gene and then see how many different patterns there are. How many different ways do humans splice this gene.
There are some settings in BLAST and BLAST download options, so you get the entire "hit" sequence in on entry rather than broken into the exons that matched your gene. But you will then have to align each of them to your gene.
Thank you everyone for your response... Since this last week, I am trying various different options and some of them have potential to get me unique hits. But as Brian mentioned, uniqueness could be at exon-intron structural level or SNPs. In my case, I want to ignore the SNPs and just go for Splice variants. And I tried Multiple sequence alignments to find out the unique patterns but the task seemed too big for the manual execution and MSA becomes really challenging when you have a lot of alternatively spliced products of all sort of sizes. In this case, it will generate unnecessary gaps or misalign the sequences.
So, kindly let me know what kind of tools or BLAST settings I shall use to get the results.
Regards
Dear Mazhar
Very interesting suggestions above. I did face a similar problem, for hypothetical protein sequences as a query though. I have always trusted Virginia FASTA instead of traditional BLAST for that the former is highly sensitive and you have an option to use FASTX Clipper to check redundant sequences. Please give it a try: fasta.bioch.virginia.edu/
Hope this helps
Prash
Dear Hussain
You can identify duplicate sequences easily in Bioedit. Open a new alignment first. Copy all sequences and import into Bioedit. Go to "Sequence" menu and select sort and Group top-down by identical sequences. It will group identical sequences together which you can remove easily.
Sincerely
K.ULAGANATHAN
Thank you everyone,
and lately thanks to Prashanth and Kandasamy for nice suggestions. I am trying the ideas suggested by all of you. I am very very sure that one or more if them will suit my need. Thank you very much again.
Yours Sincerely,
Mazhar Hussain
- Badges
- Science topic
- Similar topics
- Biological Science
- Molecular Biology
- Biomolecules
- Nucleic Acids
- DNA